CN107644364A

CN107644364A - Object filter method and system

Info

Publication number: CN107644364A
Application number: CN201710844233.3A
Authority: CN
Inventors: 王修充
Original assignee: Beijing Jingdong Century Trading Co Ltd; Beijing Jingdong Shangke Information Technology Co Ltd
Current assignee: Beijing Jingdong Century Trading Co Ltd; Beijing Jingdong Shangke Information Technology Co Ltd
Priority date: 2017-09-18
Filing date: 2017-09-18
Publication date: 2018-01-30
Also published as: WO2019052263A1

Abstract

Present disclose provides a kind of object filter method, including：Obtain the fisrt feature information of the first object and the second feature information of the second object；The similarity of the first object and the second object is determined according to the fisrt feature information of acquisition and second feature information；And judge whether similarity meets similarity threshold；If similarity meets similarity threshold, then determine that the first object and the second object are mutual filtering objects, wherein, when the first object and the second object are mutual filtering objects, if user performed assigned operation to the first object in the first preset time, user will not perform assigned operation in the second preset time for being less than preset duration with the first prefixed time interval duration to the second object.The disclosure additionally provides a kind of object filter system, a kind of computer system and a kind of computer-readable recording medium.

Description

Object filter method and system

Technical field

A kind of this disclosure relates to Internet technical field, and in particular to object filter method and system.

Background technology

With the rapid development of Internet technology and e-commerce technology, shopping online is joyous by more and more users' Meet.A commercial product recommending system part essential as ecommerce, solve problem of information overload, help user find with Purchase is adapted to commodity of oneself etc. to seem extremely important.Good commercial product recommending system can lift the sales rate of commodity and total Body turnover.

In electric business recommendation process, in order to lift Consumer's Experience, while improve the conversion ratio of recommendation, multiple purchase filtering gradually into For an important process.So-called multiple purchase filtering refers to predict based on user's commodity purchasing behavior in a short time using some way The user commodity that will not be bought again interior for a period of time are filtered, avoid causing shadow to the commodity that user recommendation needs filter Ring Consumer's Experience and the conversion ratio recommended.

For example, most of users for buying " air-conditioning " in the recent period, consider to buy the possibility of " air-conditioning " again in the short time It is smaller, if still largely recommending " air-conditioning " in recommendation, not only bad for the conversion of recommendation, and Consumer's Experience can be influenceed, Therefore need to filter out such commodity；And most of users for buying " toilet paper " in the recent period, consider to buy again in the short time The possibility of " toilet paper " is larger, if still largely recommending " toilet paper " in recommendation, is not only beneficial to the conversion recommended, and Consumer's Experience can be improved, therefore such commodity should not be filtered out.

At present, one kind is provided in the related art and purchases filtering scheme again, however, realizing the process of present inventive concept In, inventor has found following defect at least be present in correlation technique：Easily filtering is excessively during multiple purchase filtering or filtering is insufficient.

The content of the invention

In view of this, can be determined the need for filtering by judging the similarity of commodity present disclose provides a kind of, In the object filter method and its system for lifting Consumer's Experience and improving recommendation conversion ratio.

An aspect of this disclosure provides a kind of object filter method, including：Obtain the fisrt feature letter of the first object The second feature information of breath and the second object；Determined according to the above-mentioned fisrt feature information of acquisition and above-mentioned second feature information State the similarity of the first object and above-mentioned second object；And judge whether above-mentioned similarity meets similarity threshold；It is if above-mentioned Similarity meets similarity threshold, it is determined that above-mentioned first object and above-mentioned second object are mutual filtering objects, wherein, when upper When to state the first object and above-mentioned second object be mutual filtering object, if user holds in the first preset time to above-mentioned first object Went assigned operation, then above-mentioned user is being less than the second preset time of preset duration with above-mentioned first prefixed time interval duration Above-mentioned assigned operation will not be performed to above-mentioned second object.

In accordance with an embodiment of the present disclosure, if above-mentioned similarity meets similarity threshold, it is determined that above-mentioned first object and upper It is that mutual filtering object includes to state the second object：If satisfied, then according to the above-mentioned fisrt feature information of acquisition and above-mentioned second spy Reference breath determines above-mentioned first object and above-mentioned second object classification region residing in vector space；Judge the class determined Whether mesh region belongs to default classification region；And if it is determined that classification region belong to default classification region, it is determined that it is above-mentioned First object and above-mentioned second object are mutual filtering objects.

In accordance with an embodiment of the present disclosure, the above method also includes：If it is determined that classification region be not belonging to default classification area Domain, it is determined that above-mentioned first object and above-mentioned second object are not mutual filtering objects.

In accordance with an embodiment of the present disclosure, the fisrt feature information of the first object and the second feature information of the second object are obtained Including：Obtain the key words text of all objects in the first category corresponding to above-mentioned first object；Determine in above-mentioned first category The distributed expression of the key words text of all objects；To the distribution of the key words text of all objects in above-mentioned first category Represent to be averaging, obtain the first eigenvector of the above-mentioned fisrt feature information of above-mentioned first object；Obtain above-mentioned second object The key words text of all objects in corresponding second category；Determine the key words text of all objects in above-mentioned second category Distribution represents；And the distributed of the key words text of all objects in above-mentioned second category is represented to be averaging, obtain State the second feature vector of the above-mentioned second feature information of the second object.

In accordance with an embodiment of the present disclosure, determined according to the above-mentioned fisrt feature information of acquisition and above-mentioned second feature information Stating the similarity of the first object and above-mentioned second object includes：By the first eigenvector of above-mentioned fisrt feature information and above-mentioned the The second feature vector of two characteristic informations makes the difference, and obtains cross feature vector；And/or ask above-mentioned first eigenvector and above-mentioned the The COS distance of two characteristic vectors；It is and/or above-mentioned according to above-mentioned first eigenvector and the vector determination of above-mentioned second feature respectively The classification attribute of first object and above-mentioned second object；And according to above-mentioned cross feature is vectorial and/or above-mentioned COS distance and/ Or above-mentioned classification attribute determines the similarity of above-mentioned first object and above-mentioned second object.

In accordance with an embodiment of the present disclosure, judge whether above-mentioned similarity meets that similarity threshold includes：To above-mentioned similarity Given a mark, obtain corresponding similarity score；Above-mentioned similarity score is inputted into two grader set in advance, so that above-mentioned Two grader output category results；Judge whether above-mentioned classification results meet the first preset value, wherein, if above-mentioned classification results etc. In above-mentioned first preset value, it is determined that above-mentioned similarity meets above-mentioned similarity threshold, if above-mentioned classification results are not equal to above-mentioned First preset value, it is determined that above-mentioned similarity is unsatisfactory for above-mentioned similarity threshold.

Another aspect of the disclosure provides a kind of object filter system, including acquisition module, for obtaining first pair The fisrt feature information of elephant and the second feature information of the second object；First determining module, for above-mentioned first according to acquisition Characteristic information and above-mentioned second feature information determine the similarity of above-mentioned first object and above-mentioned second object；And first judge Module, for judging whether above-mentioned similarity meets similarity threshold；Second determining module, for meeting phase in above-mentioned similarity In the case of like degree threshold value, determine that above-mentioned first object and above-mentioned second object are mutual filtering objects, wherein, when above-mentioned first When object and above-mentioned second object are mutual filtering objects, refer to if user performed in the first preset time to above-mentioned first object Fixed operation, then above-mentioned user will not be right in the second preset time for being less than preset duration with above-mentioned first prefixed time interval duration Above-mentioned second object performs above-mentioned assigned operation.

In accordance with an embodiment of the present disclosure, above-mentioned second determining module includes：First determining unit, in above-mentioned similarity In the case of meeting similarity threshold, above-mentioned is determined according to the above-mentioned fisrt feature information of acquisition and above-mentioned second feature information The classification region of one object and above-mentioned second object residing in vector space；First judging unit, for judging what is determined Whether classification region belongs to default classification region；And second determining unit, for belonging to default in the classification region determined In the case of classification region, determine that above-mentioned first object and above-mentioned second object are mutual filtering objects.

In accordance with an embodiment of the present disclosure, said system also includes：3rd determining module, in the classification region determined In the case of being not belonging to default classification region, determine that above-mentioned first object and above-mentioned second object are not mutual filtering objects.

In accordance with an embodiment of the present disclosure, above-mentioned acquisition module includes：First acquisition unit, for obtaining above-mentioned first object The key words text of all objects in corresponding first category；3rd determining unit, for determining own in above-mentioned first category The distributed expression of the key words text of object；First computing unit, for the key to all objects in above-mentioned first category The distributed of word text represents to be averaging, and obtains the first eigenvector of the above-mentioned fisrt feature information of above-mentioned first object；The Two acquiring units, for obtaining the key words text of all objects in the second category corresponding to above-mentioned second object；4th determines Unit, for determining that the distributed of the key words text of all objects represents in above-mentioned second category；And second computing unit, For representing to be averaging to the distributed of the key words text of all objects in above-mentioned second category, above-mentioned second object is obtained The second feature vector of above-mentioned second feature information.

In accordance with an embodiment of the present disclosure, above-mentioned first determining module includes：3rd computing unit, for special by above-mentioned first The first eigenvector and the second feature vector of above-mentioned second feature information of reference breath make the difference, and obtain cross feature vector；With/ Or the 4th computing unit, for seeking the COS distance of above-mentioned first eigenvector and above-mentioned second feature vector；And/or the 5th is true Order member, for determining above-mentioned first object and above-mentioned the according to above-mentioned first eigenvector and above-mentioned second feature vector respectively The classification attribute of two objects；And the 6th determining unit, for according to above-mentioned cross feature is vectorial and/or above-mentioned COS distance And/or above-mentioned classification attribute determines the similarity of above-mentioned first object and above-mentioned second object.

In accordance with an embodiment of the present disclosure, above-mentioned first judge module includes：Marking unit, for being carried out to above-mentioned similarity Marking, obtains corresponding similarity score；Processing unit, for above-mentioned similarity score input set in advance two to be classified Device, so that above-mentioned two graders output category result；Second judging unit, for judging whether above-mentioned classification results meet first Preset value, wherein, if above-mentioned classification results are equal to above-mentioned first preset value, it is determined that above-mentioned similarity meets above-mentioned similarity threshold Value, if above-mentioned classification results are not equal to above-mentioned first preset value, it is determined that above-mentioned similarity is unsatisfactory for above-mentioned similarity threshold.

Another aspect of the present disclosure provides a kind of computer system, including：One or more processors；Memory, use In storing one or more programs, wherein, when said one or multiple programs are by said one or multiple computing devices, make Said one or multiple processors realize object filter method any one of above-described embodiment.

Another aspect of the present disclosure provides a kind of computer-readable recording medium, is stored thereon with executable instruction, should Instruction makes processor realize the object filter method any one of above-described embodiment when being executed by processor.

In accordance with an embodiment of the present disclosure, because employing by judging that the similarity of commodity determines the need for being filtered The technological means of operation, so at least partially overcoming, easily filtering is excessively when correlation technique purchases filtering again or filtering is not filled Point technical problem, and then lifting Consumer's Experience and the technique effect for improving the filtering degree of accuracy.

Brief description of the drawings

By the description to the embodiment of the present disclosure referring to the drawings, the above-mentioned and other purposes of the disclosure, feature and Advantage will be apparent from, in the accompanying drawings：

Fig. 1 diagrammatically illustrate according to the embodiment of the present disclosure can be with application filter method and the system tray of system Structure；

Fig. 2 diagrammatically illustrates the flow chart of the object filter method according to the embodiment of the present disclosure；

Fig. 3 A diagrammatically illustrate determines first pair when similarity meets similarity threshold according to the embodiment of the present disclosure Flow chart as with the second object being mutual filtering object；

Fig. 3 B diagrammatically illustrate the flow chart of the object filter method according to another embodiment of the disclosure；

Fig. 3 C diagrammatically illustrate the object of acquisition first according to the embodiment of the present disclosure fisrt feature information and second pair The flow chart of the second feature information of elephant；

Fig. 3 D diagrammatically illustrate to be believed according to the embodiment of the present disclosure according to the fisrt feature information and second feature of acquisition Breath determines the flow chart of the similarity of the first object and the second object；

Fig. 3 E diagrammatically illustrate judges whether similarity meets the flow of similarity threshold according to the embodiment of the present disclosure Figure；

Fig. 4 diagrammatically illustrates the block diagram of the object filter system according to the embodiment of the present disclosure；

Fig. 5 A diagrammatically illustrate the block diagram of the second determining module according to the embodiment of the present disclosure；

Fig. 5 B diagrammatically illustrate the block diagram of the object filter system according to another embodiment of the disclosure；

Fig. 5 C diagrammatically illustrate the block diagram of the acquisition module according to the embodiment of the present disclosure；

Fig. 5 D diagrammatically illustrate the block diagram of the first determining module according to the embodiment of the present disclosure；

Fig. 5 E diagrammatically illustrate the block diagram of the first judge module according to the embodiment of the present disclosure；And

Fig. 6 diagrammatically illustrates the frame of the computer system for being adapted for carrying out object filter method according to the embodiment of the present disclosure Figure.

Embodiment

Hereinafter, it will be described with reference to the accompanying drawings embodiment of the disclosure.However, it should be understood that these descriptions are simply exemplary , and it is not intended to limit the scope of the present disclosure.In addition, in the following description, the description to known features and technology is eliminated, with Avoid unnecessarily obscuring the concept of the disclosure.

Term as used herein is not intended to limit the disclosure just for the sake of description specific embodiment.Use herein Term " comprising ", "comprising" etc. indicate the presence of the feature, step, operation and/or part, but it is not excluded that in the presence of Or addition one or more other features, step, operation or parts.

All terms (including technology and scientific terminology) as used herein have what those skilled in the art were generally understood Implication, unless otherwise defined.It should be noted that term used herein should be interpreted that with consistent with the context of this specification Implication, without should by idealization or it is excessively mechanical in a manner of explain.

, in general should be according to this using in the case of being similar to that " in A, B and C etc. at least one " is such and stating Art personnel are generally understood that the implication of the statement to make an explanation (for example, " having system at least one in A, B and C " Should include but is not limited to individually with A, individually with B, individually with C, with A and B, with A and C, with B and C, and/or System with A, B, C etc.).Using in the case of being similar to that " in A, B or C etc. at least one " is such and stating, it is general come Say be generally understood that the implication of the statement to make an explanation (for example, " having in A, B or C at least according to those skilled in the art The system of one " should include but is not limited to individually with A, individually with B, individually with C, with A and B, with A and C, with B and C, and/or system etc. with A, B, C).It should also be understood by those skilled in the art that substantially arbitrarily represent two or more The adversative conjunction and/or phrase of optional project, either in specification, claims or accompanying drawing, shall be construed as Give including one of these projects, the possibility of these projects either one or two projects.For example, " A or B " should for phrase It is understood to include " A " or " B " or " A and B " possibility.

Embodiment of the disclosure provides a kind of for determining that filtering object is carried with this by judging the similarity of commodity Consumer's Experience is risen with improving the object filter method of the filtering degree of accuracy and the object filter system of this method can be applied.The party Method includes obtaining the fisrt feature information of the first object and the second feature information of the second object, is believed according to the fisrt feature of acquisition Breath and second feature information determine the similarity of the first object and the second object, and judge whether similarity meets similarity threshold Value, if similarity meets similarity threshold, it is determined that the first object and the second object are mutual filtering objects, wherein, when first When object and the second object are mutual filtering objects, if user performed assigned operation in the first preset time to the first object, Then user will not perform to the second object in the second preset time for being less than preset duration with the first prefixed time interval duration and refer to Fixed operation.

Fig. 1 diagrammatically illustrate according to the embodiment of the present disclosure can be with application filter method and the system tray of system Structure.

As shown in figure 1, terminal device 101,102,103, network can be included according to the system architecture 100 of the embodiment 104 and server 105.Network 104 is to the offer communication link between terminal device 101,102,103 and server 105 Medium.Network 104 can include various connection types, such as wired, wireless communication link or fiber optic cables etc..

User can be interacted with using terminal equipment 101,102,103 by network 104 with server 105, to receive or send out Send message etc..Various telecommunication customer end applications, such as the application of shopping class, net can be installed on terminal device 101,102,103 (merely illustrative) such as the application of page browsing device, searching class application, JICQ, mailbox client, social platform softwares.

Terminal device 101,102,103 can have a display screen and a various electronic equipments that supported web page browses, bag Include but be not limited to smart mobile phone, tablet personal computer, pocket computer on knee and desktop computer etc..

Server 105 can be to provide the server of various services, such as utilize terminal device 101,102,103 to user The website browsed provides the back-stage management server (merely illustrative) supported.Back-stage management server can be to the use that receives The data such as family request analyze etc. processing, and by result (such as according to user's acquisition request or the webpage of generation, believe Breath or data etc.) feed back to terminal device.

It should be noted that the object filter method that the embodiment of the present disclosure is provided can typically be performed by server 105. Correspondingly, the object filter system that the embodiment of the present disclosure is provided can be typically arranged in server 105.The embodiment of the present disclosure The object filter method provided can also by different from server 105 and can with terminal device 101,102,103 and/or clothes The server or server cluster that business device 105 communicates perform.Correspondingly, the object filter system that the embodiment of the present disclosure is provided It can be arranged at different from server 105 and the service that can be communicated with terminal device 101,102,103 and/or server 105 In device or server cluster.

It should be understood that the number of the terminal device, network and server in Fig. 1 is only schematical.According to realizing need Will, can have any number of terminal device, network and server.

Fig. 2 diagrammatically illustrates the flow chart of object filter method in accordance with an embodiment of the present disclosure.

As shown in Fig. 2 this method can include operation S201~S204, wherein：

In operation S201, the fisrt feature information of the first object and the second feature information of the second object are obtained.

It should be noted that the first object and the second object can represent commodity, specifically, the first object can represent Main commodity, corresponding fisrt feature information can be expressed as the main commodity in itself possessed characteristic information (title of such as commodity, Operational factor, energy-conservation attribute etc.)；Second object can represent that the candidate's commodity filtered, the expression of second feature information may be needed For candidate's commodity possessed characteristic information in itself.

In embodiment of the disclosure, the form of expression of characteristic information can include a variety of, will not be repeated here.For example, Characteristic information can be represented with characteristic vector or feature distribution formula.

For example, user have purchased a " frequency conversion ", " micro- intelligence ", " the Gree sky of " energy-conservation " on certain electric business platform Adjust ", then the first object can be that " air-conditioning 1 ", corresponding fisrt feature information can be " Gree ", " frequency conversion ", " micro- intelligence Energy ", " energy-conservation "；Accordingly, for a candidate's commodity that may need to filter, such as " fixed frequency ", " intelligence arc ", " light sensing " " beautiful air-conditioning ", the second object can be " air-conditioning 2 ", corresponding second feature information can be " beautiful air-conditioning ", " fixed frequency ", " intelligence arc ", " light sensing ".Further, " characteristic information " Gree " of air-conditioning 1 ", " frequency conversion ", " micro- intelligence are extracted respectively Energy ", " energy-conservation ", and " characteristic information " beautiful air-conditioning " of air-conditioning 2 ", " fixed frequency ", " intelligence arc ", " light sensing ".

In operation S202, the first object and the second object are determined according to the fisrt feature information of acquisition and second feature information Similarity.

In the embodiment originally opened, determine the similarity of the first object and the second object determination mode can include it is more Kind, it will not be repeated here, for example, at least (feature vector, X 1 and second of the first object can be included by independent characteristic vector The feature vector, X 2 of object) or at least one of is combined by independent characteristic vector determined：Cross feature vector (i.e. two The difference (X1-X2) of individual independent characteristic vector), X1 and X2 COS distance, classification attribute (its of the first object and the second object In, if the classification attribute of two objects is identical, 1 can be expressed as, is otherwise expressed as 0) to determine.Wherein, independent spy is passed through Sign vector combines cross feature vector and/or when COS distance and/or classification attribute determines, they can be spliced into as follows to Measure (X1, X2, R), R represents at least one in cross feature vector, COS distance and classification attribute, it should be appreciated that if R is represented Cross feature vector, COS distance and any two in classification attribute or three, then show as splicing form.For example, R is represented Cross feature vector sum COS distance, then R show as the splicing form of cross feature vector sum COS distance.

In operation S203, judge whether similarity meets similarity threshold.

In embodiment of the disclosure, judge whether similarity meets that similarity threshold can include various ways, herein Repeat no more.For example, being given a mark to similarity, obtain similarity score, further, judge similarity score whether be more than or Person is equal to similarity score threshold value set in advance.

For example, similarity score threshold value is set as 0.5, if the similarity score of the first object and the second object is 0.7, by In 0.7 ＞ 0.5, therefore show that the first object and the second object are similar；If the similarity score of the first object and the second object is 0.3, due to 0.3 ＜ 0.5, therefore an object and the second object are dissimilar.

In operation S204, if similarity meets similarity threshold, it is determined that the first object and the second object are to filter mutually Object.Wherein, when the first object and the second object are mutual filtering objects, if user in the first preset time to the first object Assigned operation was performed, then user will not be right in the second preset time for being less than preset duration with the first prefixed time interval duration Second object performs assigned operation.

It should be noted that filtering object is expressed as mutually, meet in the similarity of the first object and the second object similar When spending threshold value, object that the second object is defined as filtering relative to the needs of the first object.Grasped according in the embodiment of the present disclosure Make the example in S201, if " air-conditioning 1 " and " similarity of air-conditioning 2 " meets similarity threshold, will " air-conditioning 2 " is defined as relatively In " the object that the needs of air-conditioning 1 " filter.Conversely, if similarity is unsatisfactory for similarity threshold, " air-conditioning 1 " and " air-conditioning 2 " is not It is mutual filtering object.

If for example, user have purchased in May, 2017 " air-conditioning 1 ", user in 1 year from May, 2017, " possibility of air-conditioning 2 " typically understands very little, is needed " air-conditioning 2 " filters out when now doing commercial product recommending for purchase again.

In accordance with an embodiment of the present disclosure, by judging the similarity of main commodity and candidate's commodity, to determine that candidate's commodity are No is the commodity that needs filter, and easily causes to filter insufficient when overcoming multiple purchase filtering in correlation technique or filters excessively Problem.

Below with reference to Fig. 3 A~Fig. 3 E, the method shown in Fig. 2 is described further in conjunction with specific embodiments.

Fig. 3 A diagrammatically illustrate the flow chart of the object filter method according to another embodiment of the disclosure.

Because although similarity is high between some commodity, may actually be not appropriate for as mutual filtering pair As, for example, for belonging to the commodity of general consumption product, it is general unsuitable as mutual filtering object, now, in commodity If filtered out either one as the mutual filtering object of the opposing party in recommendation process, then filtering can be caused excessive, in order to enter The problem of one step overcomes filtering excessive, the disclosure additionally provides a kind of optional embodiment.In this embodiment, the object filter Method can also include operation S301~S303 in addition to it can include above with reference to operation S201~S204 of Fig. 2 descriptions. For description for purpose of brevity, the description to operating S201~S204 in Fig. 2 is omitted here.As shown in Figure 3A, wherein：

S301 is being operated, if similarity meets similarity threshold, according to the fisrt feature information and second feature of acquisition Information determines the first object and the second object classification region residing in vector space；

In operation S302, whether the classification region for judging to determine belongs to default classification region；And

In operation S303, however, it is determined that the classification region gone out belongs to default classification region, it is determined that the first object and second pair As if mutual filtering object.

For example, in the case of being all " air-conditioning class " commodity for the first object and the second object, both sentence by similarity When breaking to be mutual filtering object, in order to prevent that filtering is excessive, the classification region belonging to both is may further determine that, due to two Person belongs to air-conditioning classification region, and the commodity under this classification region can be mutual filtering object.For another example for first In the case that object and the second object are all " hygienic stationery " commodity, both are judged as YES mutual filtering object by similarity When, in order to prevent that filtering is excessive, the classification region belonging to both is may further determine that, because both belong to toilet paper classification Region, and the commodity under this classification region can typically be handled not as mutual filtering object.

If for example, user have purchased in May, 2017 " air-conditioning 1 ", user in 1 year from May, 2017, " possibility of air-conditioning 2 " typically understands very little, is needed " air-conditioning 2 " filters out when now doing commercial product recommending for purchase again；But if User have purchased that " toilet paper 1 ", user buy that " probability of toilet paper 2 " is still very big in a short time in July, 2017.In order to Avoid " toilet paper 2 " is as " the mutual filtering object of toilet paper 1 " filters out, and after judging by similarity, then judges two Classification region belonging to person, finally to determine whether to regard both as mutual filtering object.

It should be noted that default classification region representation is the classification region that pre-set needs are filtered.

In embodiment of the disclosure, the simple similarity by judging the first object and the second object, to determine to be It is no to need to filter candidate's commodity, it can cause to filter excessive phenomenon generation.Such as " facial mask class " commodity, the face in vector space Film classification region, although similarity is big, belong to running stores, in a short time, user repeatedly buys the probability of " facial mask class " commodity Still very big, therefore, it is the commodity that need not be filtered, i.e. facial mask classification that " facial mask class " commodity, which can be trained to better category of model, Region is not belonging to default classification region.Again such as " air-conditioning 1 " and " air-conditioning 2 " is " air-conditioning class " commodity, in air-conditioning classification region, User may be difficult the commodity for buying the classification region again in a short time, and therefore, " air-conditioning class " commodity can be by being trained to Category of model to need the object that filters, i.e. air-conditioning classification region belongs to default classification region, in other words, " air-conditioning 1 " and " empty It is mutual filtering object to adjust 2 ".

In accordance with an embodiment of the present disclosure, it is determined that the similarity of main commodity and candidate's commodity meets the condition of similarity threshold Under, judge main commodity and the classification region residing for candidate's commodity, if to need the default classification region filtered, further Determine whether main commodity and candidate's commodity are mutual filtering object, filter the problem of excessive so as to further overcome, reach The effect of lifting Consumer's Experience.

Fig. 3 B diagrammatically illustrate the flow chart of the object filter method according to another embodiment of the disclosure.

In this embodiment, the object filter method is except that can include the operation above with reference to Fig. 2 and Fig. 3 A descriptions Outside S201~S204 and S301~S302, operation S401 (operate S303 and replace with operation S401) can also be included.In order to Description for purpose of brevity, omits the description to operating S201~S204 and S301~S302 in Fig. 2 and Fig. 3 A here.Such as Fig. 3 B institutes Show, wherein：

In operation S401, however, it is determined that the classification region gone out is not belonging to default classification region, it is determined that the first object and second Object is not mutual filtering object.

For example, " facial mask 1 " and " the facial mask classification region that facial mask 2 " be in space vector, be not belonging to preset classification region, Then " facial mask 1 " and " facial mask 2 " is not mutual filtering object.

In accordance with an embodiment of the present disclosure, by judging main commodity and the classification region residing for candidate's commodity, it is not belonging to preset Classification region, determine whether that main commodity and candidate's commodity are not mutual filtering objects, it is filtered so as to further avoid The generation of phenomenon is spent, reaches raising and recommends quality, and increase the technique effect for recommending conversion ratio.

Fig. 3 C diagrammatically illustrate the object of acquisition first according to the embodiment of the present disclosure fisrt feature information and second pair The flow chart of the second feature information of elephant.

In this embodiment, the object filter method except can include above with reference to Fig. 2 description operation S202~ Outside S204, operation S501~S506 can also be included (operation S501~S506 can be included by operating S201).In order to describe For purpose of brevity, the description to operating S202~S204 in Fig. 2 is omitted here.As shown in Figure 3 C, wherein：

The key words text of all objects in operation S501, the first category corresponding to the first object of acquisition；

In operation S502, the distributed expression of the key words text of all objects in the first category is determined；

In operation S503, the distributed of the key words text of all objects in the first category is represented to be averaging, obtain the The first eigenvector of the fisrt feature information of one object；

The key words text of all objects in operation S504, the second category corresponding to the second object of acquisition；

In operation S505, the distributed expression of the key words text of all objects in the second category is determined；And

In operation S506, the distributed of the key words text of all objects in the second category is represented to be averaging, obtain the The second feature vector of the second feature information of two objects.

It should be noted that the first category or the second category can be " air-conditioning class ", " facial mask class ", " hygienic stationery ", This is not limited.

In embodiment of the disclosure, it is first determined the affiliated category of the first object, select all objects in the category Key words text, carry out the unsupervised training of distributed expression.Represented for the distribution of the first object, key can be shown as The vector form of word text.Further, the distribution is represented to be averaging, you can obtain the characteristic vector (i.e. the of the first object One characteristic vector).

For example, for the first object, " air-conditioning 1 ", its first category are expressed as " air-conditioning class ", obtain in " air-conditioning class " and own Air-conditioning (as " air-conditioning 1 ", " air-conditioning 2 " ... " and air-conditioning n ") key words text." if air-conditioning class " commodity only include " air-conditioning 1 " " air-conditioning 2 ", then the distribution of " air-conditioning class " be expressed as [Gree, frequency conversion, micro- intelligence, energy-conservation, beautiful air-conditioning, fixed frequency, Intelligence arc, light sensing].Another step, the distributed of " air-conditioning class " is represented to be averaging, that is, obtain first object " air-conditioning 1 " Characteristic vector.

Likewise, according to the above method, the characteristic vector (i.e. second feature vector) of the second object can be obtained, herein not Repeat again.

In accordance with an embodiment of the present disclosure, the method represented using distribution, can effectively retain commodity key words text, In addition, for the commodity under same classification region, the characteristic vector after being represented with distribution, it will in vector space Same panel region in, so not only remain the key words texts of commodity, while be also easy to use very much machine learning classification Device is divided to vector space, and whether the vector space filtered in needs is divided according to main commodity and candidate's commodity Class.Meanwhile distribution is represented to be averaging, further increase the accuracy of filtering.

Fig. 3 D diagrammatically illustrate to be believed according to the embodiment of the present disclosure according to the fisrt feature information and second feature of acquisition Breath determines the flow chart of the similarity of the first object and the second object.

In this embodiment, the object filter method is except that can include the operation above with reference to Fig. 2 and Fig. 3 C descriptions Outside S203, S204 and S501~S506, can also including operation S601~S604, (operation S601 can be included by operating S202 ~S604).For description for purpose of brevity, omit and retouched to operating S203, S204 and S501~S506 in Fig. 2 and Fig. 3 C here State.As shown in Figure 3 D, wherein：

In operation S601, the first eigenvector of fisrt feature information and the second feature vector of second feature information are done Difference, obtain cross feature vector；And/or

In operation S602, the first eigenvector COS distance vectorial with second feature is sought；And/or

In operation S603, the first object and the second object are determined according to first eigenvector and second feature vector respectively Classification attribute；And

In operation S604, the first object and the are determined according to cross feature vector and/or COS distance and/or classification attribute The similarity of two objects.

In embodiment of the disclosure, determine similarity mode have it is a variety of, preferably according to the first object and second pair The cross feature vector of elephant, COS distance and the triplicity of classification attribute determine similarity, do not limit herein.Specifically, Determine that similarity can be classified by gradient boosted tree (Gradient Boosting Decision Tree, referred to as GBDT) Device is realized.

For example, judging similarity by COS distance, when COS distance is smaller, the first object and the second object are represented Similarity it is smaller, if similarity now is unsatisfactory for similarity threshold, the first object and the second object are not to filter mutually Object, it is not necessary to carry out filter operation.But when COS distance is larger, represent that similarity is larger, further, it is also necessary to sentence The classification region that disconnected first object and the second object are in, if belong to default classification region, and then determine the first object and the Whether two objects are mutual filtering object.

It should be noted that determine that similarity is a machine learning method for having supervision, now, in the training set of machine learning Training sample be changed into ((X, X1, R), label), wherein, sku is expressed as the first object, and sku1 is expressed as the second object, X tables The characteristic vector of the first object is shown as, X1 is expressed as the characteristic vector of the second object, and it is vectorial and/or remaining that R is expressed as cross feature Chordal distance and/or classification attribute, label are expressed as whether the second object needs to filter.

In accordance with an embodiment of the present disclosure, similarity is judged for the characteristic vector of main commodity and candidate's commodity, further The degree of accuracy of filtering is added, meanwhile, improve the experience effect of user.

Fig. 3 E are diagrammatically illustrated according to the embodiment of the present disclosure for judging whether similarity meets similarity threshold Flow chart.

In this embodiment, the object filter method is except that can include operation S201, S202 above with reference to Fig. 2 descriptions Outside S204, operation S701~S703 can also be included (operation S701~S703 can be included by operating S203).In order to retouch State for purpose of brevity, omit the description to operating S201, S202 and S204 here.As shown in FIGURE 3 E, wherein：

In operation S701, similarity is given a mark, obtains corresponding similarity score；

In operation S702, similarity score is inputted into two grader set in advance, so that two grader output category knots Fruit；And

In operation S703, judge whether classification results meet the first preset value.Wherein, preset if classification results are equal to first Value, it is determined that similarity meets similarity threshold, if classification results are not equal to the first preset value, it is determined that similarity is unsatisfactory for phase Like degree threshold value.

, can be by scoring functions L (sku, sku1)=score to the first object and second in embodiment of the disclosure The similarity of object is given a mark, wherein, score is expressed as similarity score, score ∈ [0,1].As score=0, table Bright first object and the second object are two entirely different commodity, it is not necessary to are filtered；As score=1, show the first object With the second object be same commodity, it is necessary to filter；When score ∈ (0,1), can be judged using two graders (this is the mode of a machine learning for having supervision, and training set is { (sku, sku1) label }), in score>, will when=0.5 The commodity that second object tag filters for needs, now, label=1, in score<It is to be not required to by the second object tag when 0.5 The commodity to be filtered, now, label=0.

Specifically, when being given a mark using scoring functions to similarity, can be realized by GBDT graders, can be with LR (Logistics Regression, logistic regression), SVM (Support Vector Machine, SVMs) or RF (Random Forest, random forest) is realized, is not limited herein.

For example, using all samples in training set { (sku, ku1), label } to be trained using GBDT graders, learn Acquistion is to a scoring functions L (sku, sku1).GBDT is one and creates multiple weak typing decision trees by iteration optimization come structure Into the sorting algorithm of final strong categorised decision tree.In GBDT iteration, it is assumed that the strong learner that previous round iteration obtains is f_t-1(x), loss function is L (y, f_t-1(x)), then the target of epicycle iteration is to find a CART trees (Classification And Regression Tree, post-class processing) model weak learner h_t(x) loss L (y, the f of epicycle, are allowed_t-1(x))=L (y,f_t(x)+h_t(x) it is) minimum.That is, the decision tree that often wheel iteration is found, will allow the loss of sample to become smaller as far as possible.

For binary GBDT sorting algorithms, using the logarithm loss function of similar logistic regression, then loss function is：

L (y, f (x))=log (1+exp (- yf (x))) y ∈ { -1 ,+1 }

Now calculate negative gradient error and the optimal residual error match value of each leaf node of the decision tree of generation.

Wherein negative gradient error is：

r_tj=y_i/(1+exp(yf(x_i)))

The optimal residual error of each leaf node is fitted to：

Strong learner is updated afterwards：

Obtain the expression formula of final strong learner：

The score values of scoring functions L (sku, skul) outputs in this work, are as the GBDT strong classifiers in above formula Expression formula f (x) result, wherein, the combinations of features feature that the x in f (x) is sku and sku1_sku,skul, i.e. L (sku, Skul)=f (featuresku, skul).The scoring functions filtered so whether will be just needed to sku and sku1, be converted into by After sku and sku1 features are combined, the classification scoring functions that are obtained using the classifier trainings of GBDT bis-.

In accordance with an embodiment of the present disclosure, similarity is given a mark using two graders, and then judges whether similarity is full Sufficient similarity threshold, the accuracy of judgement is improved, optimize commercial product recommending effect.

Fig. 4 diagrammatically illustrates the block diagram of the object filter system according to the embodiment of the present disclosure.

In this embodiment, the object filter system 400 can include acquisition module 410, the first determining module 420, the One judge module 430 and the second determining module 440.The object filter system 400 can be performed above with reference to Fig. 2, Fig. 3 A~figure The method of 3E descriptions.As shown in figure 4, wherein：Acquisition module 410, for obtaining the fisrt feature information and second of the first object The second feature information of object, the first determining module 420 are true for the fisrt feature information according to acquisition and second feature information The similarity of fixed first object and the second object, and the first judge module 430, for judging whether similarity meets similarity Threshold value, the second determining module 440, in the case of meeting similarity threshold in similarity, determine the first object and second pair As if mutual filtering object, wherein, when the first object and the second object are mutual filtering objects, if user is when first is default Between assigned operation was performed to the first object, then user is being less than the second pre- of preset duration with the first prefixed time interval duration If the time will not perform assigned operation to the second object.

Fig. 5 A diagrammatically illustrate the block diagram of the second determining module according to the embodiment of the present disclosure.

In this embodiment, the object filter system 400 except can include above with reference to Fig. 4 description corresponding module it Outside, the second determining module 440 can also include the first determining unit 441, the first judging unit 442 and the second determining unit 443. For description for purpose of brevity, the description to corresponding module in Fig. 4 is omitted here.As shown in Figure 5A, wherein：First determining unit 441, in the case of being unsatisfactory for similarity threshold in similarity, according to the fisrt feature information and second feature information of acquisition Determine the first object and the second object classification region residing in vector space, the first judging unit 442, for judging to determine Whether the classification region gone out belongs to default classification region, and the second determining unit 443, in the classification region category determined In the case of default classification region, determine that the first object and the second object are mutual filtering objects.

Fig. 5 B diagrammatically illustrate the block diagram of the object filter system according to another embodiment of the disclosure.

In this embodiment, the object filter system 400 is except that can include above with reference to the corresponding of Fig. 4 and Fig. 5 A descriptions Outside module, the 3rd determining module 510 can also be included.For description for purpose of brevity, omit here to phase in Fig. 4 and Fig. 5 A Answer the description of module.As shown in Figure 5 B, wherein：3rd determining module 510, for being not belonging to preset in the classification region determined In the case of classification region, determine that the first object and the second object are not mutual filtering objects.

Fig. 5 C diagrammatically illustrate the block diagram of the acquisition module according to the embodiment of the present disclosure.

In this embodiment, the object filter system 400 except can include above with reference to Fig. 4 description corresponding module it Outside, acquisition module 410 can also include first acquisition unit 411, the 3rd determining unit 412, the first computing unit 413, second Acquiring unit 414, the 4th determining unit 415 and the second computing unit 416.For description for purpose of brevity, omit here to Fig. 4 With the description of corresponding module in Fig. 5 A.As shown in Figure 5 C, wherein：First acquisition unit 411, for obtaining corresponding to the first object The key words text of all objects in first category, the first determining unit 412, for determining the pass of all objects in the first category The distributed expression of keyword text, the first computing unit 413, for point to the key words text of all objects in the first category Cloth represents to be averaging, and obtains the first eigenvector of the fisrt feature information of the first object, second acquisition unit 414, is used for Obtain the key words text of all objects in the second category corresponding to the second object, the second determining unit 415, for determining second The distributed expression of the key words text of all objects in category, and, the second computing unit 416, in the second category The distributed of the key words text of all objects represents to be averaging, obtain the second object second feature information second feature to Amount.

Fig. 5 D diagrammatically illustrate the block diagram of the first determining module according to the embodiment of the present disclosure.

In this embodiment, the object filter system 400 is except that can include above with reference to the corresponding of Fig. 4 and Fig. 5 C descriptions Outside module, the first determining module 420 can also include the 3rd computing unit 421, the 4th computing unit the 422, the 5th determines list The determining unit 424 of member 423 and the 6th.For description for purpose of brevity, omit here and corresponding module in Fig. 4 and Fig. 5 C is retouched State.As shown in Figure 5 D, wherein：3rd computing unit 421, for by the first eigenvector and second feature of fisrt feature information The second feature vector of information makes the difference, and obtains cross feature vector, and/or, the 4th computing unit 422, for seeking fisrt feature The COS distance of vector and second feature vector, and/or, the 5th determining unit 423, for respectively according to first eigenvector and Second feature vector determines the classification attribute of the first object and the second object, and, the 6th determining unit 425, for according to friendship Fork characteristic vector and/or COS distance and/or classification attribute determine the similarity of the first object and the second object.

Fig. 5 E diagrammatically illustrate the block diagram of the first judge module according to the embodiment of the present disclosure.

In this embodiment, the object filter system 400 except can include above with reference to Fig. 4 description corresponding module it Outside, the first judge module 430 can also include marking unit 431, processing unit 432 and judging unit 433.For the letter of description For the sake of clean, the description to corresponding module in Fig. 4 is omitted here.As shown in fig. 5e, wherein：Marking unit 431, for similarity Given a mark, obtain corresponding similarity score, processing unit 432, for similarity score to be inputted into two points set in advance Class device, so that two grader output category results, the second judging unit 433, for judging whether classification results meet that first is pre- If value, wherein, if classification results are equal to the first preset value, it is determined that similarity meets similarity threshold, if classification results In the first preset value, it is determined that similarity is unsatisfactory for similarity threshold.

It is understood that acquisition module 410, the first determining module 420, the first judge module 430, the second determining module 440 and the 3rd determining module 510 may be incorporated in a module and realize, or any one module therein can be split It is divided into multiple modules.Or at least part function of one or more of these modules module can be with other modules extremely Small part function phase combines, and is realized in a module.According to an embodiment of the invention, acquisition module 410, first determines mould At least one in block 420, the first judge module 430, the second determining module 440 and the 3rd determining module 510 can be at least Be implemented partly as on hardware circuit, such as field programmable gate array (FPGA), programmable logic array (PLA), piece be System, the system on substrate, the system in encapsulation, application specific integrated circuit (ASIC), or can be to be integrated or be encapsulated to circuit The hardware such as any other rational method or firmware realize, or with software, three kinds of implementations of hardware and firmware it is suitable Realized when combination.Or acquisition module 410, the first determining module 420, the first judge module 430, the second determining module 440 And the 3rd at least one in determining module 510 can at least be implemented partly as computer program module, when the program When being run by computer, the function of corresponding module can be performed.

Fig. 6 diagrammatically illustrates the frame of the computer system for being adapted for carrying out object filter method according to the embodiment of the present disclosure Figure.Computer system shown in Fig. 6 is only an example, the function and use range of the embodiment of the present disclosure should not be brought and appointed What is limited.

As shown in fig. 6, including processor 601 according to the computer system 600 of the embodiment of the present disclosure, it can be according to storage Program in read-only storage (ROM) 602 is loaded into random access storage device (RAM) 603 from storage part 1008 Program and perform various appropriate actions and processing.Processor 601 for example can include general purpose microprocessor (such as CPU), Instruction set processor and/or related chip group and/or special microprocessor (for example, application specific integrated circuit (ASIC)), etc..Place Reason device 601 can also include being used for the onboard storage device for caching purposes.Processor 601 can include being used to perform with reference to figure 2, figure Single treatment unit either multiple places of the different actions of the method flow according to the embodiment of the present disclosure of 3A~Fig. 3 E descriptions Manage unit.

In RAM 603, it is stored with computer system 600 and operates required various programs and data.Processor 601, ROM 602 and RAM 603 is connected with each other by bus 604.Processor 601 is by performing the journey in ROM 602 and/or RAM 603 Sequence performs the various operations described above with reference to Fig. 2, Fig. 3 A~Fig. 3 E.Removed it is noted that described program can also be stored in In one or more memories beyond ROM 602 and RAM 603.Processor 601 can also be stored in described one by performing Program in individual or multiple memories performs the various operations described above with reference to Fig. 2, Fig. 3 A~Fig. 3 E.

In accordance with an embodiment of the present disclosure, computer system 600 can also include input/output (I/O) interface 605, input/ Output (I/O) interface 605 is also connected to bus 604.Computer system 600 can also include being connected to the following of I/O interfaces 605 It is one or more in part：Importation 606 including keyboard, mouse etc.；Including such as cathode-ray tube (CRT), liquid crystal The output par, c 607 of display (LCD) etc. and loudspeaker etc.；Storage part 608 including hard disk etc.；And including such as The communications portion 609 of the NIC of LAN card, modem etc..Communications portion 609 is held via the network of such as internet Row communication process.Driver 610 is also according to needing to be connected to I/O interfaces 605.Detachable media 611, such as disk, CD, magnetic CD, semiconductor memory etc., it is arranged on as needed on driver 610, in order to the computer program read from it Storage part 608 is mounted into as needed.

In accordance with an embodiment of the present disclosure, it may be implemented as computer software journey above with reference to the method for flow chart description Sequence.For example, embodiment of the disclosure includes a kind of computer program product, it includes carrying meter on a computer-readable medium Calculation machine program, the computer program include the program code for being used for the method shown in execution flow chart.In such embodiments, The computer program can be downloaded and installed by communications portion 609 from network, and/or be pacified from detachable media 611 Dress.When the computer program is performed by processor 601, the above-mentioned function of being limited in the system of the embodiment of the present disclosure is performed.Root According to embodiment of the disclosure, system as described above, unit, module, unit etc. can by computer program module come Realize.

It should be noted that the computer-readable medium shown in the disclosure can be computer-readable signal media or meter Calculation machine readable storage medium storing program for executing either the two any combination.Computer-readable recording medium for example can be --- but not Be limited to --- electricity, magnetic, optical, electromagnetic, system, device or the device of infrared ray or semiconductor, or it is any more than combination.Meter The more specifically example of calculation machine readable storage medium storing program for executing can include but is not limited to：Electrical connection with one or more wires, just Take formula computer disk, hard disk, random access storage device (RAM), read-only storage (ROM), erasable type and may be programmed read-only storage Device (EPROM or flash memory), optical fiber, portable compact disc read-only storage (CD-ROM), light storage device, magnetic memory device, Or above-mentioned any appropriate combination.In the disclosure, computer-readable recording medium can any include or store journey The tangible medium of sequence, the program can be commanded the either device use or in connection of execution system, device.And at this In open, computer-readable signal media can be included in a base band or the data-signal as carrier wave part propagation, Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can Any computer-readable medium beyond storage medium is read, the computer-readable medium, which can send, propagates or transmit, to be used for By instruction execution system, device either device use or program in connection.Included on computer-readable medium Program code can be transmitted with any appropriate medium, be included but is not limited to：Wirelessly, electric wire, optical cable, RF etc., or it is above-mentioned Any appropriate combination.In accordance with an embodiment of the present disclosure, computer-readable medium can include above-described ROM 602 And/or one or more memories beyond RAM 603 and/or ROM 602 and RAM 603.

Flow chart and block diagram in accompanying drawing, it is illustrated that according to the system of the various embodiments of the disclosure, method and computer journey Architectural framework in the cards, function and the operation of sequence product.At this point, each square frame in flow chart or block diagram can generation The part of one module of table, program segment or code, a part for above-mentioned module, program segment or code include one or more For realizing the executable instruction of defined logic function.It should also be noted that some as replace realization in, institute in square frame The function of mark can also be with different from the order marked in accompanying drawing generation.For example, two square frames succeedingly represented are actual On can perform substantially in parallel, they can also be performed in the opposite order sometimes, and this is depending on involved function.Also It is noted that the combination of each square frame and block diagram in block diagram or flow chart or the square frame in flow chart, can use and perform rule Fixed function or the special hardware based system of operation are realized, or can use the group of specialized hardware and computer instruction Close to realize.

As on the other hand, the disclosure additionally provides a kind of computer-readable medium, is stored thereon with executable instruction, should Instruction makes processor realize the object filter method any one of the above method embodiment when being executed by processor.The calculating Machine computer-readable recording medium can be included in the equipment described in above-described embodiment；Can also be individualism, and without supplying In the equipment.Above computer computer-readable recording medium carries one or more program, when said one or multiple programs are by one When the individual equipment performs so that the equipment performs：Obtain the fisrt feature information of the first object and the second feature of the second object Information；The similarity of the first object and the second object is determined according to the fisrt feature information of acquisition and second feature information；And Judge whether similarity meets similarity threshold；If similarity meets similarity threshold, it is determined that the first object and the second object It is mutual filtering object, wherein, when the first object and the second object are mutual filtering objects, if user is in the first preset time Assigned operation was performed to the first object, then user presets with second of the first prefixed time interval duration less than preset duration Time will not perform assigned operation to the second object.

Embodiment of the disclosure is described above.But the purpose that these embodiments are merely to illustrate that, and It is not intended to limit the scope of the present disclosure.Although respectively describing each embodiment more than, but it is not intended that each reality Use can not be advantageously combined by applying the measure in example.The scope of the present disclosure is defined by the appended claims and the equivalents thereof.Do not take off From the scope of the present disclosure, those skilled in the art can make a variety of alternatives and modifications, and these alternatives and modifications should all fall at this Within scope of disclosure.

Claims

1. a kind of object filter method, including：

Obtain the fisrt feature information of the first object and the second feature information of the second object；

First object and described second pair is determined according to the fisrt feature information of acquisition and the second feature information The similarity of elephant；And

Judge whether the similarity meets similarity threshold；

If satisfied, then determine that first object and second object are mutual filtering objects, wherein, when first object When with second object being mutual filtering object, if user performed in the first preset time to first object specifies behaviour Make, then the user will not be to described in the second preset time for being less than preset duration with the first prefixed time interval duration Second object performs the assigned operation.

2. the method according to claim 11, wherein, if satisfied, then determining that first object and second object are Mutual filtering object includes：

If satisfied, first object and institute are then determined according to the fisrt feature information of acquisition and the second feature information State the second object classification region residing in vector space；

Whether the classification region for judging to determine belongs to default classification region；And

If belong to, it is determined that first object and second object are mutual filtering objects.

3. according to the method for claim 2, wherein, methods described also includes：

If it is not belonging to, it is determined that first object and second object are not mutual filtering objects.

4. the method according to claim 11, wherein, the fisrt feature information of the first object of acquisition and the second of the second object Characteristic information includes：

Obtain the key words text of all objects in the first category corresponding to first object；

Determine the distributed expression of the key words text of all objects in first category；

The distributed of the key words text of all objects in first category is represented to be averaging, obtains first object The first eigenvector of the fisrt feature information；

Obtain the key words text of all objects in the second category corresponding to second object；

Determine the distributed expression of the key words text of all objects in second category；And

The distributed of the key words text of all objects in second category is represented to be averaging, obtains second object The second feature vector of the second feature information.

5. according to the method for claim 4, wherein, believed according to the fisrt feature information of acquisition and the second feature Breath determines that the similarity of first object and second object includes：

The first eigenvector of the fisrt feature information and the second feature vector of the second feature information are made the difference, obtained Cross feature vector；And/or

Seek the COS distance of the first eigenvector and second feature vector；And/or

First object and second object are determined according to the first eigenvector and second feature vector respectively Classification attribute；And

According to the cross feature is vectorial and/or the COS distance and/or the classification attribute determine first object and The similarity of second object.

6. according to the method for claim 1, wherein, judge whether the similarity meets that similarity threshold includes：

The similarity is given a mark, obtains corresponding similarity score；

The similarity score is inputted into two grader set in advance, so that the two graders output category result；

Judge whether the classification results meet the first preset value, wherein, if the classification results are equal to first preset value, Then determine that the similarity meets the similarity threshold, if the classification results are not equal to first preset value, it is determined that The similarity is unsatisfactory for the similarity threshold.

7. a kind of object filter system, including：

Acquisition module, for obtaining the fisrt feature information of the first object and the second feature information of the second object；

First determining module, described first is determined for the fisrt feature information according to acquisition and the second feature information The similarity of object and second object；And

First judge module, for judging whether the similarity meets similarity threshold；

Second determining module, in the case of satisfaction, determining that first object and second object are to filter mutually Object, wherein, when first object and second object are mutual filtering objects, if user is in the first preset time pair First object performed assigned operation, then the user is being less than preset duration with the first prefixed time interval duration The second preset time will not perform the assigned operation to second object.

8. system according to claim 7, wherein, second determining module includes：

First determining unit, in the case of satisfaction, according to the fisrt feature information of acquisition and the second feature Information determines first object and second object classification region residing in vector space；

Whether the first judging unit, the classification region for judging to determine belong to default classification region；And

Second determining unit, in the case where belonging to, determining that first object and second object are to filter mutually Object.

9. system according to claim 8, wherein, the system also includes：

3rd determining module, in the case where being not belonging to, determining that first object and second object are not mutual Filtering object.

10. system according to claim 7, wherein, the acquisition module includes：

First acquisition unit, for obtaining the key words text of all objects in the first category corresponding to first object；

3rd determining unit, for determining that the distributed of the key words text of all objects represents in first category；

First computing unit, for representing to be averaging to the distributed of the key words text of all objects in first category, Obtain the first eigenvector of the fisrt feature information of first object；

Second acquisition unit, for obtaining the key words text of all objects in the second category corresponding to second object；

4th determining unit, for determining that the distributed of the key words text of all objects represents in second category；And

Second computing unit, for representing to be averaging to the distributed of the key words text of all objects in second category, Obtain the second feature vector of the second feature information of second object.

11. system according to claim 10, wherein, first determining module includes：

3rd computing unit, for by the second of the first eigenvector of the fisrt feature information and the second feature information Characteristic vector makes the difference, and obtains cross feature vector；And/or

4th computing unit, for seeking the COS distance of the first eigenvector and second feature vector；And/or

5th determining unit, for determining described first pair according to the first eigenvector and second feature vector respectively As the classification attribute with second object；And

6th determining unit, for according to the cross feature is vectorial and/or the COS distance and/or the classification attribute it is true The similarity of fixed first object and second object.

12. system according to claim 7, wherein, first judge module includes：

Marking unit, for being given a mark to the similarity, obtains corresponding similarity score；

Processing unit, for the similarity score to be inputted into two grader set in advance, so that two grader exports Classification results；

Second judging unit, for judging whether the classification results meet the first preset value, wherein, if described classification results etc. In first preset value, it is determined that the similarity meets the similarity threshold, if the classification results are not equal to described First preset value, it is determined that the similarity is unsatisfactory for the similarity threshold.

13. a kind of computer system, including：

One or more processors；

Memory, for storing one or more programs,

Wherein, when one or more of programs are by one or more of computing devices so that one or more of Processor realizes the object filter method any one of claim 1 to 6.

14. a kind of computer-readable recording medium, is stored thereon with executable instruction, the instruction makes processing when being executed by processor Device realizes the object filter method any one of claim 1 to 6.